CN102469978B - Noise reduction of breathing signals - Google Patents
Noise reduction of breathing signals Download PDFInfo
- Publication number
- CN102469978B CN102469978B CN201080030613.XA CN201080030613A CN102469978B CN 102469978 B CN102469978 B CN 102469978B CN 201080030613 A CN201080030613 A CN 201080030613A CN 102469978 B CN102469978 B CN 102469978B
- Authority
- CN
- China
- Prior art keywords
- frequency
- time
- signal
- breathing
- gain function
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
- 230000029058 respiratory gaseous exchange Effects 0.000 title claims abstract description 125
- 230000009467 reduction Effects 0.000 title claims abstract description 48
- 238000001914 filtration Methods 0.000 claims abstract description 63
- 238000000034 method Methods 0.000 claims abstract description 30
- 230000003595 spectral effect Effects 0.000 claims abstract description 25
- 238000001228 spectrum Methods 0.000 claims description 87
- 238000010183 spectrum analysis Methods 0.000 claims description 7
- 230000006978 adaptation Effects 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 238000012545 processing Methods 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 49
- 238000010586 diagram Methods 0.000 description 10
- 238000012360 testing method Methods 0.000 description 10
- 208000037656 Respiratory Sounds Diseases 0.000 description 8
- 238000005070 sampling Methods 0.000 description 8
- 230000003860 sleep quality Effects 0.000 description 6
- 230000015572 biosynthetic process Effects 0.000 description 5
- 238000003786 synthesis reaction Methods 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 238000004590 computer program Methods 0.000 description 4
- 230000000241 respiratory effect Effects 0.000 description 4
- 230000003044 adaptive effect Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 230000002349 favourable effect Effects 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 2
- 230000000052 comparative effect Effects 0.000 description 2
- 210000003437 trachea Anatomy 0.000 description 2
- 206010041235 Snoring Diseases 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 239000004568 cement Substances 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000004064 recycling Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000009131 signaling function Effects 0.000 description 1
- 238000011410 subtraction method Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
- 230000026676 system process Effects 0.000 description 1
- 230000002618 waking effect Effects 0.000 description 1
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B7/00—Instruments for auscultation
- A61B7/003—Detecting lung or respiration noise
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/08—Detecting, measuring or recording devices for evaluating the respiratory organs
- A61B5/0803—Recording apparatus specially adapted therefor
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0208—Noise filtering
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B2562/00—Details of sensors; Constructional details of sensor housings or probes; Accessories for sensors
- A61B2562/02—Details of sensors specially adapted for in-vivo measurements
- A61B2562/0204—Acoustic sensors
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61M—DEVICES FOR INTRODUCING MEDIA INTO, OR ONTO, THE BODY; DEVICES FOR TRANSDUCING BODY MEDIA OR FOR TAKING MEDIA FROM THE BODY; DEVICES FOR PRODUCING OR ENDING SLEEP OR STUPOR
- A61M2230/00—Measuring parameters of the user
- A61M2230/40—Respiratory characteristics
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Veterinary Medicine (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Pulmonology (AREA)
- Physics & Mathematics (AREA)
- Physiology (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Computational Linguistics (AREA)
- Quality & Reliability (AREA)
- Signal Processing (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
- Soundproofing, Sound Blocking, And Sound Damping (AREA)
Abstract
The invention relates to a system for and a method of processing breathing signals. A noise reduction operation is performed on a spectral breathing signal (18) to compute an output spectral signal (38), said noise reduction operation using spectral subtraction; and a two-dimensional frequency and time filtering (32) of a gain function (30) used in the spectral subtraction of the noise reduction operation performing step is performed, for example, a two-dimensional frequency and time median filtering of the gain function. For example, said spectral breathing signal is computed based on a breathing signal.
Description
Technical field
The present invention relates to process breathing (breathing) signal (especially sound breathing signals) field, relate more particularly to the process method of breathing signals and the system for the treatment of breathing signals.
Background technology
Breathing signals or breathing (respiratory) signal and breathing rate are basic vital signs.Breathing rate can such as obtain from the respiratory waveform measured.Such as, respiratory waveform signal produces by being attached to the sensor electrode of the people that will measure its breathing from outside.Respiratory waveform also can be derived from electrocardiogram (ECG) waveform.
US6290654B1 discloses a kind of equipment of the breathing pattern for detecting breathing patient.Antenna Type mike is used for sensing the sound of breathing or snoring.The acoustical signal listened from patient body converts the signal of telecommunication to by this mike and another trachea mike be positioned on patient's cervical region, and these signals of telecommunication are provided to A/D converter.Output from corresponding A/D transducer is provided to active noise and eliminates unit, this active noise eliminates unit for using the adaptive line wave filter Background suppression noise be made up of one group of delay line components, and each delay line components is represented by a delay sampling cycle and one group of corresponding scalable coefficient.The output of adaptive line wave filter is deducted from sensor exports.The tap-weights that the output obtained is used in adjustment adaptive line wave filter is to minimize the mean-square value of total output.Signal noise from trachea mike processes in the mode similar with the signal noise from Antenna Type mike.Signal is through bandpass filtering and be supplied to the volume of the estimation of airflow waveform generating unit.
Summary of the invention
Need to place the measurement of sensor electrode or mike being attached to people be actually offend the eye and be inconvenient for people.This is even more important when processing the breathing signals of people of sleep.
Therefore, the breathing signals of the microphones capture of being placed by the people away from breathing can desirably be processed.
Such as, when mike record breathing signals, mike can be placed in the vicinity of the people of breathing, such as, away from the people 50cm of breathing.But because the acoustic energy caused by breathing is usually very weak, thus signal to noise ratio, the ratio namely between breath signal and noise may be very low, makes the related breathing parameter extracting such as breathing rate and so on from signal be difficult.
Desirably can perform the noise reduction operation of new kind, to be conducive to determining respiration parameter when catching sound breathing signals from people a distance of sleep.
Have been found that most of mike (it has the amplifier be built in mike) shows such sensor noise, this sensor noise shows low-pass characteristic due to the 1/f noise of amplifier.Some are as MEMS(MEMS) mike sensor has the signal to noise ratio (snr) of approximate 60dB, i.e. signal-sensor noise ratio.Other more excellent mikes may have the SNR scope of 70-80dB.
Desirably provide a kind of method for the treatment of breathing signals or system, described method or system are conducive to reducing the sensor noise from mike.
In order to solve better these close hit one or more, in a first aspect of the present invention, provide a kind of method processing breathing signals, the method comprises:
-noise reduction operation is performed to calculate output spectrum signal for frequency spectrum breathing signals, described noise reduction operation uses spectral subtraction; And
-perform two-dimensional frequency and time filtering that noise reduction operation performs the gain function used in the spectral subtraction of step.
Such as, noise reduction operation comprises the step of the two-dimensional filtering of described execution gain function.
Such as, described method comprises further:
-based on breathing signals, such as, based on the breathing signals in time domain, calculate described frequency spectrum breathing signals.
Such as, breathing signals is sound breathing signals.Such as, breathing signals is by the signal of microphones capture.
Such as, the step calculating frequency spectrum breathing signals comprises fast Fourier transform (FFT).
Term " noise reduction operation " is appreciated that and represents the applicable operation reducing the noise (such as sensor noise) comprised in the signal represented by frequency spectrum breathing signals.
Term " filtering " is appreciated that and comprises average, medium filtering and other filtering algorithms.
In the noise reduction operation using spectral subtraction, as described in further detail hereinafter further, use gain function.
Usually, breathing slowly increases along with the time and reduces.In addition, the frequency content of breathing signals shows significant peak and paddy unlike such as in the formant of voice signal.Have been found that the spectral characteristic of breathing sound shows the relevant information up to approximate 2kHz.Huge peak and the large regions of paddy is not shown owing to existing in the time-frequency characteristic of breathing signals, thus can the advantageously two-dimensional frequency of using gain function and time filtering.Therefore, noise reduction can be improved while the unacceptable distortion avoiding breathing signals.
Such as, the described two-dimensional frequency of gain function and time filtering are two-dimensional frequency and time average/medium filtering, i.e. average or medium filtering.Usually, the filtering algorithm as average or medium filtering is particularly suitable for eliminating exceptional value (outlier).
In one embodiment, the two-dimensional frequency of gain function and time filtering are two-dimensional frequency and time medium filtering.Medium filtering is particularly advantageous, because it allows by replacing exceptional value by other values, unsmooth relevant signal section assigns to eliminate exceptional value.Therefore, the quality of noise reduction operation is improved.
Such as, described method comprises further based on output spectrum signal is transformed into time domain and provides output signal.This may be favourable for calculating respiration parameter.Such as, output spectrum signal is transformed into time domain and can comprises inverse fast fourier transform (IFFT).
Such as, described method comprises and calculates at least one respiration parameter based on output spectrum signal.An important respiration parameter is such as breathing rate.Such as, at least one respiration parameter described is calculated based on the described output signal obtained by output spectrum signal is transformed into time domain.Such as, at least one respiration parameter described is sleep quality parameter.Therefore, sleep quality can be determined.Such as, can determine or estimate the sleep quality of the people slept, and can estimate to provide control signal based on sleep quality.Such as, this control signal can be the signal of the external factor (such as temperature, illumination etc.) for controlling or affect sleep environment.Such as, described one or more control signal can be selected for the sleep of the people improving or strengthen sleep.Another kind of probability is when the sleep quality of people is estimated as too low (such as lower than sleep quality threshold value), is provided for the control signal waking people up.
In one embodiment, calculate based on the noise estimated and amplitude spectrum signal the gain function used in spectral subtraction, described amplitude spectrum signal is calculated based on frequency spectrum breathing signals.Such as, calculated gains function can comprise the noise spectrum of estimation divided by absolute value spectrum signal.But, replace absolute value spectrum signal, also can use such as squared absolute value, i.e. power spectrum signal.Therefore, in this case, based on the noise calculation gain function that the quadratic sum of amplitude spectrum signal is estimated.
Such as, noise reduction operation comprises step:
-calculate amplitude spectrum signal based on frequency spectrum breathing signals;
-based on the noise estimated and amplitude spectrum signal calculated gains function;
The two-dimensional frequency of-execution gain function and time filtering; And
-calculate output spectrum signal by gain function frequency spectrum breathing signals being multiplied by filtering.
In other words, noise reduction operation comprises the step of the two-dimensional filtering of above-mentioned execution gain function.Therefore, efficient noise reduction operation is achieved.
Such as, noise reduction operation comprises estimating noise, and described estimating noise comprises the average operation of the some continuous blocks based on amplitude spectrum signal, and described amplitude spectrum signal is calculated based on frequency spectrum breathing signals.Such as, can be averaged to amplitude spectrum signal on some continuous blocks.Alternatively, such as, can be averaged to square (i.e. a power spectrum signal) for amplitude spectrum signal.Therefore, the stable priori of noise is used.Particularly advantageous in this situation that such as noise is determined primarily of the sensor noise of mike wherein, because such sensor noise is highly stable.Such as, for each frequency slots (bin), namely define each element of the discrete frequency group of spectrum signal function thereon, the e.g. particularly frequency slots of FFT, perform average operation.Such as, the quantity of described continuous blocks is corresponding to the time period of at least 1 second, preferably at least 3 seconds, especially at least 10 seconds.Therefore, high-quality Noise Background can be realized estimate.
Such as, the described two-dimensional frequency and the time filtering that perform gain function at least 0.05 second, more preferably at least 0.1 second or filter kernel that even more preferably the time span of at least 0.25 second is corresponding is utilized.
Such as, utilize the filter kernel corresponding at least 3 continuous blocks of the amplitude spectrum signal used in noise reduction operation to perform described two-dimensional frequency and the time filtering of gain function, block is of a size of such as 512 samples.More preferably, the quantity of block is at least 5 or even more preferably at least 7.Therefore, noise reduction can be improved.
Such as, utilize at least 40Hz, more preferably at least 75Hz or filter kernel that even more preferably at least the frequency span of 100Hz is corresponding perform described two-dimensional frequency and the time filtering of gain function.Such as, filter kernel can be corresponding to the frequency span of at least 250Hz.Such as, the described two-dimensional frequency and the time filtering that perform gain function at least 3 frequency slots, more preferably at least 5 frequency slots or filter kernel that even more preferably at least 7 frequency slots are corresponding is utilized.Therefore, sensor noise can be removed efficiently.
Such as, described method comprises with the sample frequency sampling breathing signals of at least 4kHz, preferably at least 8kHz.The use of this relatively high sample frequency is favourable, because the spectral characteristic of breathing sound shows the relevant information up to 2kHz.Therefore, noise reduction can be improved.
Such as, described method can comprise the output signal that down-sampling calculates based on output spectrum signal.Use and allow to improve noise reduction for noise reduction operation than calculating the sample frequency that needed for respiration parameter, frequency is higher.Such as, sample frequency can be at least 8kHz.In addition, high sample frequency is used to allow to perform gain function two-dimensional filtering over frequency and over time and do not make breathing audio distortions.
In one embodiment, described method comprises filter kernel width and/or the filter kernel length of the filter kernel used in the described two-dimensional frequency of adapt gain function and time filtering further; Described adapt is based on the ratio according to the two-dimensional frequency of gain function and time filtering and between the noise energy (energy of such as music tone) comprised in the signal energy comprised in the first time-frequency region of the signal obtained and the second time-frequency region.Such signal is the gain function of such as filtering, frequency spectrum output function or output function, and it all depends on the gain function of filtering.Such as, described signal can be by the two-dimensional frequency of gain function and time filtering and the gain function of the filtering obtained.Such as, the first time-frequency region is made up of first frequency scope and very first time scope, and the second time-frequency region is made up of second frequency scope and the second time range.Such as, first frequency scope is lower frequency range, and second frequency scope is higher frequency range.Filter kernel width is corresponding to the frequency span of filter kernel, and filter kernel length is corresponding to the time span of filter kernel.Preferably, the first and second time-frequency region are selected such that the energy of the signal (i.e. breathing sound) wanted produces sizable contribution for the energy comprised in the first time-frequency region.Therefore, but the energy comprised in neighborhood in the first time-frequency region outside, the energy comprised in the second time-frequency region especially suitably selected can be summed up as noise, such as music tone.
Term " signal energy " and " noise energy " refer to the integrated square of the absolute value of signal amplitude, and can in time domain or frequency-domain calculations.In other words, term " noise energy " is appreciated that and represents that supposition can be summed up as the signal energy of noise.
Such as, described adapt comprise optimize described function the first time-frequency region in ratio between the noise energy that comprises in the signal energy that comprises and the second time-frequency region.
Such as, very first time scope can be selected to make it corresponding to or comprise (locally) peak signal energy of described signal.
Such as, described first and second time range can not have overlap.Such as, described first and second time range can be continuous print time range.Such as, the second time range can before very first time scope.
Such as, step by step, such as, according to the frequency slots of fixed qty or the time block adapt filter kernel width of fixed qty and/or filter kernel length.
Such as, described signal is obtained based on the two-dimensional frequency of gain function and time filtering, described filtering utilizes test filter core to perform, and the width of this test filter core and/or length are different from width and/or the length of the filter kernel of current use in the step of the filtering performing gain function respectively.
For following reason, adapt filter kernel width and/or filter kernel length are favourable.
The characteristic time frequency pattern of breathing signals can change along with time and concrete people.Such as, people can have air-breathing and expiration, and wherein most of breathing concentration of energy is in specific time span, and for other people, breathing energy may scatter across longer time span.Similarly, frequency characteristic also can change along with different people.Except varying with each individual, also may there is the difference of time-frequency characteristic along with the time of the breathing signals of single people.Such as, when this people catches a cold, usual breathing process is occurred by mouth instead of nose, and time-frequency characteristic probably changes.
Although the music tone that spectral subtraction causes can by selecting enough large core size and minimize or be removed, too large core size may cause the distortion of breathing sound, such as frequency spectrum and/or temporally to smear (smear out).Therefore, adapt filter kernel width and/or filter kernel length can improve noise reduction while the distortion minimizing breathing sound.
In another aspect of this invention, provide a kind of system for the treatment of breathing signals, this system comprises:
-noise reduction unit, it is for calculating output spectrum signal by performing noise reduction operation to frequency spectrum breathing signals, and wherein noise reduction operation uses spectral subtraction; And
-filter cell or wave filter, it is for performing two-dimensional frequency and the time filtering of the gain function used in spectral subtraction.
Such as, described system can be suitable for the method performing process breathing signals as described above.
Such as, described system comprises further:
-spectral analysis unit or frequency spectrum analyser, it is for based on breathing signals, and the breathing signals namely in time domain, calculates described frequency spectrum breathing signals.
Such as, breathing signals is input to spectral analysis unit.Such as, the output comprising frequency spectrum breathing signals of spectral analysis unit is input to noise reduction unit.
Such as, output spectrum signal is exported by noise reduction unit.Such as, noise reduction unit comprises filter cell.Such as, filter cell is the median filter unit of two-dimensional frequency for performing gain function and time medium filtering.Such as, spectral analysis unit, noise reduction unit and/or filter cell are the part of the spectral subtraction unit for the system to breathing signals filtering.
Such as, described system comprises further for calculating the synthesis unit of output signal based on output spectrum signal is transformed into time domain.
Such as, the input of noise reduction unit is coupled in the output of spectral analysis unit.Such as, the input of synthesis unit is coupled in the output of noise reduction unit.
Such as, described system comprises the breathing analytic unit for calculating at least one respiration parameter based on output spectrum signal further.Such as, the output of the output of noise reduction unit or the output of spectral subtraction unit or synthesis unit can be coupled to the input of breathing analytic unit.
In one embodiment, described system comprises filter adaptation regulon further, the filter kernel width of the filter kernel used in its described two-dimensional frequency for adapt gain function and time filtering and/or filter kernel length, wherein said adapt is based on the ratio according to the two-dimensional frequency of gain function and time filtering and between the noise energy comprised in the signal energy comprised in the first time-frequency region of the signal obtained and the second time-frequency region.Filter adaptation regulon can be arranged for the step performing adapt filter kernel width as described above and/or filter kernel length.
These and other aspects of the present invention will be well-known according to embodiment described below, and set forth with reference to these embodiments.
Accompanying drawing explanation
Fig. 1 schematically shows for the treatment of the data flow in the system of breathing signals and this system.
Fig. 2 show graphic extension breathe air-breathing signal, the output signal of comparative example and the diagram of acoustic energy of output signal of Fig. 1 system.
Fig. 3 schematically shows the setting of the breathing signals of the people for obtaining sleep.
Fig. 4 schematically shows another embodiment of the system for the treatment of breathing signals.
Detailed description of the invention
Fig. 1 schematically illustrates the system for the treatment of breathing signals.System and method for the treatment of breathing signals is described hereinafter with reference to Fig. 1.
In FIG, signal or data flow is schematically shown by arrow.Thin-line arrow illustrates the signal in time domain, if or explicitly to mention, one or more parameter is shown.The thick arrow of profile that has represents the data block stream in time domain or the data block stream in frequency domain.
The breathing signals 10 being placed in the microphones capture of the distance of the people such as 50cm from sleep is sampled by sampler or sampling unit 12 and is arranged to time block.Such as, sample frequency is 8kHz.The sample rate of 8kHz is enough for the relevant information of the breathing sound kept in breathing signals.Such as, continuous print B sample block is converted the signal into.Such as, B=256.
The input of block cascade and windowing unit 14 is coupled in the output of sampling unit 12, and this block cascade and windowing unit are by constructing continuous print time block or M sample block by the sample of current block and the sample cascade from least one previous block.Such as, the sample of the sample of current block and previous block is cascaded into M sample, wherein M=2B.Such as, M=512.
Such as, M sample of block is by the windowing of Hann window, and Hann window is also referred to as raised cosine window.
The input of spectral analysis unit 16 is coupled in the output of unit 14.Such as, unit 16 uses FFT computing that M sample block is transformed into frequency domain.Therefore, unit 16 export have the complex-specturm of M frequency slots continuous overlapping time block form frequency spectrum breathing signals 18.The input of amplitude spectrum computing unit 20 is coupled in the output of unit 16.The complex frequency spectrum of spectrum signal 18 is converted to the amplitude spectrum signal 22 with M/2+1 frequency slots by unit 20.Such as, under the frequency resolution of 15.625Hz, M/2+1=257.
The input of noise estimation unit 24 and the input of gain function computing unit 26 are coupled in the output of unit 20.
Noise estimation unit 24 estimated spectral Noise Background and amplitude noise spectrum signal 28 is outputted to gain function computing unit 26.Therefore, another input of unit 26 is coupled in the output of unit 24.Such as, by being averaged and estimating noise to the block some continuous times of amplitude spectrum signal 22.Such as, the number of blocks of signal 22 can be corresponding to the time period of 10 seconds or more.
Unit 26 is based on amplitude spectrum signal 22 and the amplitude noise spectrum signal 28 calculated gains function 30 estimated.Such as, gain function 30 has the form of the continuous blocks of such frequency spectrum and the block size of M/2+1, and this frequency spectrum has the spectrum value between 0 and 1.
Calculating for the gain function of noise-spectrum subtraction method is similarly known in such as speech enhan-cement field.Usually, calculated gains function, makes the spectral subtraction of noise can pass through in a frequency domain spectrum signal 18 be multiplied by gain function 30 and perform.Such as, frequency spectrum or gain function 30 can be calculated as G
n=max (1-W
n/ | X|, λ), wherein X is frequency spectrum breathing signals 18, W is the noise spectrum signal estimated, and n=0 ..., M/2+1.Here, λ is the so-called frequency spectrum background of restriction noise reduction.If λ=0, so obtain maximum noise reduction.
The frequency spectrum that unit 26 exports or gain function 30 are input to filter unit 32, such as 2D median filter unit.Filter unit 32 such as performs two-dimensional frequency and the time medium filtering of gain function or frequency spectrum 30.Such as, filter unit 32 has medium filtering core, and this medium filtering core has the width of 7 frequency slots and the length of 7 time blocks.In medium filtering, by getting the mirror image of frequency spectrum and the expansion performed from M/2+1 frequency slots to M frequency slots.The gain function 34 of filter unit 32 output filtering.
The input of frequency spectrum multiplication unit 36 is coupled in the output of unit 32 and unit 16, and spectrum signal 18 is multiplied with the gain function 34 of filtering by this frequency spectrum multiplication unit, thus performs the spectral subtraction of the noise estimated.Therefore, output spectrum signal 38 is created.
The input of synthesis unit 40 is coupled in the output of unit 36, and output spectrum signal 38 is such as transformed into time domain by performing IFFT by this synthesis unit.
Therefore, unit 20,24,26,32 and 36 is formed and is used for performing noise reduction operation to spectrum signal 18, thus calculates the noise reduction unit of output spectrum signal 38.
The input of the overlapping and addition unit 42 of block is coupled in the output of unit 40, and this block overlap and addition unit are based on the addition calculation B sample block of the lap of the M sample block received from unit 40.Such as, described B sample block is non-overlapped continuous blocks.The block that unit 42 exports is converted to the continuous sample sequence forming output signal 46 by converter unit 44.Such as, converter unit 44 performs down-sampling.Therefore, output signal 46 and can have the sample frequency lower than the sample frequency of sampling unit 12.
Such as, the output of converter unit 44 is input to breathing signals assessment unit 48, and this breathing signals assessment unit calculates at least one respiration parameter 50 based on output signal 46.Respiration parameter can be such as breathing rate.Such as, breathing signals assessment unit 48 can calculate breathing rate in the mode be similarly known in the art according to output signal 46.
Such as, the system for the treatment of breathing signals can comprise the processor for performing computer program, and this computer program is suitable for performing above-described method.Such as, what this processor and/or computer program can comprise and/or be formed in above-described unit (such as unit 12,14,16,20,24,26,28,32,36,40,42,44 and/or 48) is one or more.Such as, this processor and/or computer program can be the parts being suitable for the computer performing above-described method.
Fig. 2 shows the diagram (A) represented along with the acoustic energy of the signal of time.Acoustic energy illustrates with the arbitrary unit of the signal value that the sampling unit 12 relating to Fig. 1 system exports.
Diagram (A) shows the true record of people across the breathing signals of approximate 24 seconds of breathing.This breathing signals is recorded by using the people be placed in from breathing to be similar to the mike at 50cm place.Sample rate is 8000Hz, and employs the value of B=256 and M=512.As seen from the figure, signal to noise ratio is poor, and the sound breathed is that eyes are sightless.
As a comparative example, diagram (B) shows and obtains from system as shown in Figure 1, but unit 32 does not perform the output signal of medium filtering.Therefore, diagram (B) shows the result of the conventional spectral subtraction of the noise of estimation.Have been found that spectral subtraction produces music tone in the output signal, these tones have the characteristic of the height random in time and frequency.Some in these music tones or exceptional value are visible in diagram (B).Such exceptional value easily may be categorized as breathing by breathing signals assessment unit or procedural error.
Diagram (C) shows the result utilizing and comprise the breathing signals of Fig. 1 system process diagram (A) of 2D median filter unit 32.7 frequency slots take advantage of the core size of 7 time blocks to be used for obtaining the output signal 46 shown in diagram (C).As seen from the figure, effectively removes music tone.Therefore, classification and the extraction of unit 48 pairs of respiration parameters can be improved.
Slowly to increase along with the time due to breathing and reduce; and because the frequency content of breathing also shows huge Feng Hegu unlike it is such as present in the vowel of voice, thus can in two dimension, perform medium filtering and not excessively make breathing audio distortions.
Fig. 3 schematically shows for using one or more mikes 52 of people 54 vicinity being placed in sleep to catch may arranging of breathing signals 10.Such as, two mikes 52 are placed in corner before bed.Two mikes at diverse location place are used to allow the change in location of the people 54 adapting to sleep.Such as, above-described method can be performed for two breathing signals 10 corresponding from different mike 52, and the breathing signals had compared with high s/n ratio can be selected to calculate at least one respiration parameter described.
Fig. 4 shows another embodiment of the system for the treatment of breathing signals, and the difference of this system and Fig. 1 system is, noise reduction unit comprises filter adaptation regulon 56 further.The input of filter adaptation regulon 56 is coupled in the output of gain function computing unit 26.The filter parameter used by filter unit 32 determined by filter adaptation regulon 56, especially filter kernel width N
fwith filter kernel length N
t.The parameter of unit 56 exports the parameters input being coupled to unit 32.
The core size corresponding with the length of some time block to the width of some frequency slots can be interpreted as the model order of filtering.Therefore, N
ffor for the model order of frequency and N
tfor the model order for the time.
In order to the filter kernel width N of the filter kernel that adapt unit 32 uses
fwith filter kernel length N
t, unit 56 performs following calculation procedure.
Unit 56 determines the very first time scope of the signal energy calculating the signal (i.e. breathing sound) wanted.Such as, determine that the signal energy of gain function reaches the detection moment t at maximum place.Such as, very first time scope be set to [t-0.25, t+0.25], wherein time value provides in seconds.
Such as, the second time range of the signal energy of calculating noise signal is determined.Such as, the second time range is set to [t-0.75, t-0.25].
Such as, then use the test filter core with the filter kernel of the current use of unit 32 with identical core size to perform two-dimensional frequency and the time filtering of gain function, thus create the signal of filtering.
Such as, calculate the signal energy comprised within the very first time and in first or the lower frequency ranges of the 0-2kHz of such as signal, and the signal energy calculated in second or the lower frequency range of the second time range and such as 2-4kHz, and calculate the ratio between these two signal energy values.
If this ratio is greater than specific threshold, so there is breathing sound in hypothesis.But, if this ratio is less than threshold value, then this means described music tone described in being present in higher frequency range.Therefore, noise reduction is insufficient.
When such as insufficient noise reduction, can gain function filtering be checked for the test filter core recycling test filter of one or more different IPs size and determine the step of described ratio, and optimum core size can be selected.Then, the core filter width N will selected
fwith core filter length N
toutput to filter unit 32.
Such as, core size or core rank can be used to be (N
t_test, N
f_test)=(N
t-1, N
f), (N
t, N
f-1), (N
t, N
f), (N
t, N
f+ 1), (N
t+ 1, N
f) test filter core determine described ratio, wherein (N
t, N
f) be current core size.Such as, the core rank with optimal ratio can be chosen as new core size.
Such as, can repeat above-describedly to determine the first and second time range and determine the step of described ratio for each time range for some breathings, and can be average on described some breathings by this ratio.
Such as, after special time in the past, such as every 10 minutes, unit 56 can reduce the model order for time and frequency, reduces such as be worth 1 by filter kernel width and length.Therefore, can application model rank along with the leakage of time.If in order to determine described ratio by the test filter core of use different size, only tested only and have the filter kernel width larger than the filter kernel of use current in unit 32 and/or the test filter core of larger filter kernel length, so this is useful especially.
Adapt filter kernel width as described above and/or filter kernel length allow to determine the optimal filter core size of independent individual, the breathing situation of this individual and/or acoustic condition.Therefore, the calculating of respiration parameter 50 can be improved.
Although illustrate and describe the present invention in described accompanying drawing and description above, such diagram and description should be considered to illustrative or exemplary, instead of restrictive.The present invention is not limited to the disclosed embodiments.
Those skilled in the art urban d evelopment protect of the present invention time, according to the research for described accompanying drawing, the disclosure and appended claims, should understand and implement the modification of disclosed embodiment.
In detail in the claims, word " comprises/comprises " element or step of not getting rid of other, and indefinite article " " does not get rid of plural number.Any Reference numeral in claims should not be regarded as the restriction to scope.
Claims (9)
1. process a method for breathing signals, comprising:
-noise reduction operation is performed to calculate output spectrum signal (38) for frequency spectrum breathing signals (18), described noise reduction operation uses spectral subtraction; And
-perform two-dimensional frequency and time filtering (32) that noise reduction operation performs the gain function (30) used in the spectral subtraction of step,
Wherein based on the noise estimated and amplitude spectrum signal calculated gains function (30), described amplitude spectrum signal is calculated based on frequency spectrum breathing signals, and utilizes the filter kernel corresponding to the frequency span of the time span of at least 0.05 second and at least 40Hz to perform described two-dimensional frequency and the time filtering of gain function.
2. the method for claim 1, wherein the described two-dimensional frequency of gain function (30) and time filtering (32) are two-dimensional frequency and time average or medium filtering.
3. method as claimed in claim 1 or 2, comprises further:
-based on breathing signals (10), calculate described frequency spectrum breathing signals (18).
4. method as claimed in claim 1 or 2, comprises further:
-calculate at least one respiration parameter (50) based on output spectrum signal (38).
5. method as claimed in claim 1 or 2, wherein noise reduction operation comprises step:
-calculate amplitude spectrum signal (22) based on frequency spectrum breathing signals (18);
-based on the noise (24) estimated and amplitude spectrum signal (22) calculated gains function (30);
The two-dimensional frequency of-execution gain function (30) and time filtering (32); And
-calculate output spectrum signal (38) by the gain function (34) frequency spectrum breathing signals (18) being multiplied by filtering.
6. method as claimed in claim 1 or 2, comprises further:
The filter kernel width of the filter kernel used in the described two-dimensional frequency of-adapt (56) gain function (30) and time filtering (32) and/or filter kernel length, the ratio between the noise energy comprised in the signal energy comprised in first time-frequency region of wherein said adapt based on the signal obtained according to the two-dimensional frequency of gain function and time filtering and the second time-frequency region.
7., for the treatment of a system for breathing signals, comprising:
-noise reduction unit (20,24,26,32,36), it is for calculating output spectrum signal (38) by performing noise reduction operation to frequency spectrum breathing signals (18), and wherein noise reduction operation uses spectral subtraction; And
-filter cell (32), it is for performing two-dimensional frequency and the time filtering of the gain function (30) used in spectral subtraction,
Wherein based on the noise estimated and amplitude spectrum signal calculated gains function (30), described amplitude spectrum signal is calculated based on frequency spectrum breathing signals, and utilizes the filter kernel corresponding to the frequency span of the time span of at least 0.05 second and at least 40Hz to perform described two-dimensional frequency and the time filtering of gain function.
8. system as claimed in claim 7, comprises further:
-spectral analysis unit (16), it is for calculating described frequency spectrum breathing signals (18) based on breathing signals (10).
9. system as claimed in claim 7 or 8, comprises further:
-filter adaptation regulon (56), the filter kernel width of the filter kernel used in its described two-dimensional frequency for adapt gain function (30) and time filtering (32) and/or filter kernel length, the ratio between the noise energy comprised in the signal energy comprised in first time-frequency region of wherein said adapt based on the signal obtained according to the two-dimensional frequency of gain function and time filtering and the second time-frequency region.
Applications Claiming Priority (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| EP09164723 | 2009-07-07 | ||
| EP09164723.0 | 2009-07-07 | ||
| PCT/IB2010/053042 WO2011004299A1 (en) | 2009-07-07 | 2010-07-02 | Noise reduction of breathing signals |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102469978A CN102469978A (en) | 2012-05-23 |
| CN102469978B true CN102469978B (en) | 2015-07-15 |
Family
ID=42731970
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201080030613.XA Expired - Fee Related CN102469978B (en) | 2009-07-07 | 2010-07-02 | Noise reduction of breathing signals |
Country Status (7)
| Country | Link |
|---|---|
| US (1) | US8834386B2 (en) |
| EP (1) | EP2451359B1 (en) |
| JP (1) | JP5648052B2 (en) |
| KR (1) | KR101715709B1 (en) |
| CN (1) | CN102469978B (en) |
| BR (1) | BR112012000273A8 (en) |
| WO (1) | WO2011004299A1 (en) |
Families Citing this family (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2012164482A1 (en) * | 2011-05-30 | 2012-12-06 | Koninklijke Philips Electronics N.V. | Apparatus and method for the detection of the body position while sleeping |
| WO2014021890A1 (en) | 2012-08-01 | 2014-02-06 | Dolby Laboratories Licensing Corporation | Percentile filtering of noise reduction gains |
| JP6432970B2 (en) * | 2013-09-27 | 2018-12-05 | 公立大学法人奈良県立医科大学 | Respiration monitoring apparatus and respiration monitoring method |
| CN105078505B (en) * | 2014-04-24 | 2018-01-23 | 重庆融海超声医学工程研究中心有限公司 | Physiological single processing method and processing unit |
| EP3229692B1 (en) * | 2014-12-12 | 2019-06-12 | Koninklijke Philips N.V. | Acoustic monitoring system, monitoring method, and monitoring computer program |
| WO2016122143A1 (en) * | 2015-01-28 | 2016-08-04 | Samsung Electronics Co., Ltd. | Method and apparatus for improving and monitoring sleep |
| US10825464B2 (en) | 2015-12-16 | 2020-11-03 | Dolby Laboratories Licensing Corporation | Suppression of breath in audio signals |
| US10339911B2 (en) | 2016-11-01 | 2019-07-02 | Stryker Corporation | Person support apparatuses with noise cancellation |
| WO2018122217A1 (en) * | 2016-12-28 | 2018-07-05 | Koninklijke Philips N.V. | Method of characterizing sleep disordered breathing |
| US10607590B2 (en) * | 2017-09-05 | 2020-03-31 | Fresenius Medical Care Holdings, Inc. | Masking noises from medical devices, including dialysis machines |
| TWI813584B (en) * | 2018-08-31 | 2023-09-01 | 逸陞有限公司 | Indoor active noise reduction system |
| JP7122225B2 (en) * | 2018-10-31 | 2022-08-19 | エア・ウォーター・バイオデザイン株式会社 | Processing device, system, processing method, and program |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1205133A (en) * | 1996-10-04 | 1999-01-13 | 摩托罗拉公司 | Filter co-processor |
| CN101443814A (en) * | 2005-11-08 | 2009-05-27 | 皇家飞利浦电子股份有限公司 | Method and system for filtering elongated features |
Family Cites Families (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US4630305A (en) * | 1985-07-01 | 1986-12-16 | Motorola, Inc. | Automatic gain selector for a noise suppression system |
| US4853903A (en) * | 1988-10-19 | 1989-08-01 | Mobil Oil Corporation | Method and apparatus for removing sinusoidal noise from seismic data |
| US5797852A (en) | 1993-02-04 | 1998-08-25 | Local Silence, Inc. | Sleep apnea screening and/or detecting apparatus and method |
| JP3484757B2 (en) * | 1994-05-13 | 2004-01-06 | ソニー株式会社 | Noise reduction method and noise section detection method for voice signal |
| US5687243A (en) * | 1995-09-29 | 1997-11-11 | Motorola, Inc. | Noise suppression apparatus and method |
| US6175602B1 (en) * | 1998-05-27 | 2001-01-16 | Telefonaktiebolaget Lm Ericsson (Publ) | Signal noise reduction by spectral subtraction using linear convolution and casual filtering |
| US6549586B2 (en) * | 1999-04-12 | 2003-04-15 | Telefonaktiebolaget L M Ericsson | System and method for dual microphone signal noise reduction using spectral subtraction |
| US6290654B1 (en) | 1998-10-08 | 2001-09-18 | Sleep Solutions, Inc. | Obstructive sleep apnea detection apparatus and method using pattern recognition |
| US6487257B1 (en) * | 1999-04-12 | 2002-11-26 | Telefonaktiebolaget L M Ericsson | Signal noise reduction by time-domain spectral subtraction using fixed filters |
| FI20025029A0 (en) * | 2002-05-29 | 2002-05-29 | Joni Kettunen | A method for obtaining reliable respiratory activity information from heart rate measurement |
| CA2454296A1 (en) * | 2003-12-29 | 2005-06-29 | Nokia Corporation | Method and device for speech enhancement in the presence of background noise |
| AU2005227775A1 (en) | 2004-04-04 | 2005-10-13 | Ben Gurion University Of The Negev Research And Development Authority | Apparatus and method for the detection of one lung intubation by monitoring sounds |
| US7742914B2 (en) * | 2005-03-07 | 2010-06-22 | Daniel A. Kosek | Audio spectral noise reduction method and apparatus |
| JP4686281B2 (en) * | 2005-07-06 | 2011-05-25 | 株式会社東芝 | Respiratory state determination device, respiratory state measurement method, and respiratory state determination program |
| WO2007026691A1 (en) * | 2005-09-02 | 2007-03-08 | Nec Corporation | Noise suppressing method and apparatus and computer program |
| EP1810618A1 (en) | 2006-01-24 | 2007-07-25 | Enel Produzione S.p.A. | System and method of processing acoustic signals for the energy characterisation of the respiratory condition |
| FI120964B (en) | 2007-02-15 | 2010-05-31 | Smart Valley Software Oy | Device and method for awakening a sleeping person at an appropriate time, connected to the waking time |
-
2010
- 2010-07-02 WO PCT/IB2010/053042 patent/WO2011004299A1/en active Application Filing
- 2010-07-02 CN CN201080030613.XA patent/CN102469978B/en not_active Expired - Fee Related
- 2010-07-02 KR KR1020127003028A patent/KR101715709B1/en active IP Right Grant
- 2010-07-02 EP EP10740324.8A patent/EP2451359B1/en not_active Not-in-force
- 2010-07-02 JP JP2012519094A patent/JP5648052B2/en not_active Expired - Fee Related
- 2010-07-02 US US13/382,304 patent/US8834386B2/en not_active Expired - Fee Related
- 2010-07-02 BR BR112012000273A patent/BR112012000273A8/en not_active IP Right Cessation
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1205133A (en) * | 1996-10-04 | 1999-01-13 | 摩托罗拉公司 | Filter co-processor |
| CN101443814A (en) * | 2005-11-08 | 2009-05-27 | 皇家飞利浦电子股份有限公司 | Method and system for filtering elongated features |
Non-Patent Citations (3)
| Title |
|---|
| BOR-SHING LIN,et al.WHEEZE RECOGNITION BASED ON 2D BILATERAL FILTERING OF SPECTROGRAM.《BIOMEDICAL ENGINEERINGAPPLICATIONS,BASIS & * |
| COMMUNICATIONS》.2006,第18卷(第3期), * |
| HUIJUN DING,et al.2D Spectrogram Filter for Single Channel Speech Enhancement.《Proceedings of the 7th WSEAS International Conference on Signal,speech and image processing 》.2007, * |
Also Published As
| Publication number | Publication date |
|---|---|
| EP2451359B1 (en) | 2017-09-06 |
| JP2012532650A (en) | 2012-12-20 |
| CN102469978A (en) | 2012-05-23 |
| BR112012000273A8 (en) | 2017-10-24 |
| KR20120034777A (en) | 2012-04-12 |
| EP2451359A1 (en) | 2012-05-16 |
| WO2011004299A1 (en) | 2011-01-13 |
| US20120157870A1 (en) | 2012-06-21 |
| US8834386B2 (en) | 2014-09-16 |
| KR101715709B1 (en) | 2017-03-13 |
| BR112012000273A2 (en) | 2016-02-23 |
| JP5648052B2 (en) | 2015-01-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102469978B (en) | Noise reduction of breathing signals | |
| CN109410977B (en) | Voice segment detection method based on MFCC similarity of EMD-Wavelet | |
| EP2457504B1 (en) | Sleep apnea syndrome examination device and program | |
| Yadollahi et al. | A robust method for heart sounds localization using lung sounds entropy | |
| CN108735225A (en) | It is a kind of based on human ear masking effect and Bayesian Estimation improvement spectrum subtract method | |
| KR20040075771A (en) | System for suppressing wind noise | |
| Pourazad et al. | Heart sound cancellation from lung sound recordings using time-frequency filtering | |
| Chang et al. | Performance evaluation and enhancement of lung sound recognition system in two real noisy environments | |
| CN101625869A (en) | Non-air conduction speech enhancement method based on wavelet-packet energy | |
| Zhang et al. | A novel wheeze detection method for wearable monitoring systems | |
| JP5614261B2 (en) | Noise suppression device, noise suppression method, and program | |
| WO2020024787A1 (en) | Method and device for suppressing musical noise | |
| CN113241089B (en) | Voice signal enhancement method and device and electronic equipment | |
| Lee et al. | Restoration of lung sound signals using a hybrid wavelet-based approach | |
| Ng et al. | Snore signal enhancement and activity detection via translation-invariant wavelet transform | |
| CN112185405B (en) | Bone conduction voice enhancement method based on differential operation and combined dictionary learning | |
| Gath et al. | Analysis of vocal tract parameters in Parkinsonian speech | |
| Tian et al. | Smart radar sensor for speech detection and enhancement | |
| EP4037549A1 (en) | Apparatus and method for snoring sound detection based on sound analysis | |
| JP2003510665A (en) | Apparatus and method for de-esser using adaptive filtering algorithm | |
| Moussavi et al. | Heart sound cancellation based on multiscale products and linear prediction | |
| CN111968627B (en) | Bone conduction voice enhancement method based on joint dictionary learning and sparse representation | |
| Ding et al. | Generalized subspace snoring signal enhancement based on noise covariance matrix estimation | |
| Jin et al. | A filter bank-based source extraction algorithm for heart sound removal in respiratory sounds | |
| Kim et al. | A new binary mask based on noise constraints for improved speech intelligibility |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20150715 Termination date: 20190702 |